Fault Location in Power Distribution Systems Using a Learning Algorithm for Multivariable Data Analysis

This paper proposes alternatives to improve the electric power service continuity using the learning algorithm for multivariable data analysis (LAMDA) classification technique to locate faults in power distribution systems. In this paper, the current and voltage waveforms measured during fault events are characterized to obtain a set of descriptors. These sets are analyzed by using the projection pursuit exploratory data analysis to obtain the best projection in the alpha* and beta* axes. Next, these projections are used as input data of five LAMDA nets which locate the fault in a power distribution system. The proposed methodology demands a minimum of investment from utilities since it only requires measurements at the distribution substation. The information used to estimate the fault location is the system configuration, line parameters, and data from recorders installed at the distribution substation.